Practical implementation issues of out-of-sequence measurements

• Published in: 2004 American Control Conference Proceedings; Volume 5 of 6 Vol. 5; p. 4749 vol.5
• Main Author: Lanzkron, P.J.
• Format: Conference Proceeding Journal Article
• Language: English
• Published: Piscataway NJ IEEE 01.01.2004; Evanston IL American Automatic Control Council
• Subjects: Applied sciences; Approximation algorithms; Computational complexity; Computer science; control theory; systems; Contracts; Control theory. Systems; Delay; Exact sciences and technology; Force measurement; Force sensors; Laboratories; Particle measurements; Sensor systems; Time measurement; Arrival time; Filter; Measurement sensor; Algorithm complexity; Buffer system; Sensor array; Computational complexity; Modeling; Damaging; Latency
• ISBN: 9780780383357; 0780383354
• ISSN: 0743-1619; 2378-5861
• DOI: 10.23919/ACC.2004.1384061

Summary form only given. Out-of-sequence measurements (OOSM) occur in multi-sensor systems in which there is latency between the reports of the sensors. Using these OOSM without damaging the overall tracking performance is critical to the performance of the overall tracking system. In particular, one would like to maintain the best track at any given time (i.e., process non-latent measurements) while providing an inexpensive method for updating the track state whenever the latent measurements appear. Approximation algorithms for updating both single filter (M. Malick et al., March 2001) (Y. Bar-Shalom et al., December 2002) and interacting multi-model (IMM) (Y. Bar-Shalom and H. Chen) have been suggested. All these algorithms make assumptions about the implementation that are not obvious. Among these assumptions, is the state of the system going into the update, which typically do not hold in a system with significant latencies, and assumptions about the system after the update which are unclear. Other issues, including gating, computational complexity, and missed detections have not been adequately addressed previously. We have implemented these algorithms in two of our production trackers and presents practical issues with their implementations. We discuss interpretation of the algorithms, gating, computational complexity and results from the implementation. We also presented a new approximation algorithm whose motivation is the gating problem.